Add your comments about the Circuit Note found at http://www.analog.com/en/circuits-from-the-lab/CN0134/vc.html.
If you need to operate from 500MHz to 4.4GHz, then you would just directly ac-couple the two devices (this is equivalent to 'no filter' result in CN-0134). The performance at the lower frequencies (~ < 1GHz) will be degraded due to the effect of odd order harmonics on the modulator LO quadrature accuracy and hence sideband suppression and EVM. One way around this is to split up the range into two bands (e.g. 500MHz to 2.0GHz and 2-4.4GHz) and use the auxiliary outputs on the ADF4350 for one band and the main outputs for the other band. In this way you could attempt to implement a harmonic filter for the lower band and just ac-couple through the higher band. This is outlined in the Common Variations section.
Divide-by-2 LO modulators tend to be easier to implement a broadband solution (see CN-0144) but the downside is that you are currently limited on our parts to a max. RF output of 2.2GHz using a LO of 4.4GHz. With the ADF4350 and ADL5385 you could generate 500MHz to 2.2GHz with just a simple differential connection, no harmonic filter needed.
I'm not really sure what kind of filter is being implemented and thereford what filter would I need to use. If I need to operate between 2.4 - 2.5 GHz, ¿Would I need to choose, for example, the filter C becouse my interest frecuency is in that range or that range is the rejection band and the main frecuency can't be just inside?
Yes, if operating between 2.4GHz and 2.5GHz, then just use Filter C.
the filter is implemented as a low pass Chebyshev, whose main purpose is to
filter out odd order harmonics (you need to get these below about -30dBc). There is also an
ac-coupling cap to block dc, so the overall response is really bandpass.